一种结合情绪信息分析的改进协同过滤方法

hrough mining the relationships between users and items (such as products, information, services) to provide personalized recommendations and assist their decision-making process, recommendation systems has recently become an important means to ease the "information overload" problem. It has been studied that mood plays a crucial role in human decision process. Therefore it may be an important contextual feature for recommender systems. This paper proposes an improved collaborative filtering method based on mood information analysis. It first builds user and item models based on mood features, propose a user similarity calculation method based on the mining of "user-mood" relationship, and finally present an N-adjacent user similarity calculation method through combining traditional user similarity calculation and the above mood-based similarity. We perform experimental comparisons of the new method and some baseline methods on the public Moviepilot dataset. The experiment shows that our approach is can alleviate the data sparsity problem as well as further improves recommendation accuracy.